Two roots for 'earth'

According to Bengtson, Basque lur, lu- 'earth' is related to PNC *lhemdɮɮwɨ 'earth'¹.  This rooy is also found as a loanword in substratal IE *lendh- 'open land, waste' (Celtic, Germanic, Balto-Slavic) and Uralic *lamti (*lamtз) 'lowland, meadow'. I supose this term was introduced by Neolithic farmers of the Linear Ceramics (LBK) culture. From Gaulish *landā 'open land' come the toponym Landes and Basque landa 'field, lot'2.

There's also a widespread Paleo-Eurasian root *dVQV 'dirt, clay' whose Basque reflexes are lohi, lot- 'dirty, mud' and zohi, zot- 'clod of earth; brick', the first with the standard treatment of the initial dental stop and the second with assibilation. 

This root was borrowed into PIE *dhigh- 'wall, fortification' (Sanskrit sa-dih- 'mound, heap, wall', Avestan para-daēza 'enclosure'³, Greek teîkhos ~ toîkhos 'wall'), which in NW languages refers to clay-like substances (e.g. English dough). But the native PIE reflex is *dheghóm 'earth'.

_________

¹ Dialectally also 'uncultivated land, desert'. 

² There is also the old collective Biscayan form landar 'desert'.
³ Borrowed into Greek parédeisos, hence English paradise.

Sea meadows

Although considered by Vascologists to be a borrowing from Latin sŏlum¹, Basque sor(h)o 'cultivated field; meadow'² is actually an Iberian loanword related to PNC *dʒʒǝlV 'plain, plateau'³ whose native counterpart is the archaic Lapurdian zar(h)o 'meadow'. This means PNC *ddʒ corresponds to an apico-alveolar sibilant /ś/ (Basque <s>) in Iberian but a lamino-alveolar /s/ (Basque <z>) in Vasconic.

This root is related to Eurasiatic *TSolV 'steppe, valley, meadow', reflected in Slavic *selo 'arable field' but which in other IE languages underwent a shift meaning to 'sea, lake'⁴: Greek hélos 'wet meadow, marsh', Old Indian sáras- 'lake, pond, pool'.
____________

¹ Borrowed into Basque zoru 'ground, floor'.
² With the Biscayan variant solo (B) 'field (prepared for sowing)'.

³ Bengtson proposed a different etymology from PNC *tʃʃHæɫu 'earth, ground, sand', which semantically doesn't fit.

⁴ For people grown up inland, large water extensions look like meadows.

Putting the pieces together

I've received several complaints from some of my fellow amateurs asking me for a complete explanation of my "system", including a table of detailed sound correspondences between Starostin's PNC and (Proto)-Basque, instead of focusing on individual etymologies.

I think it should be clear from earlier posts that Basque is far from being a homogeneous language and the study of its prehistory can't be separated from substrate loanwords in Western European languages.

I've already addressed the distinction between two main lexicon layers in Basque: standard Basque (that is, those words which comply with Late (Mitxelena's) Proto-Basque phonotatics, for example haga 'stake, pole'), and non-standard Basque (those which don't, for example tako 'block of wood, wedge'). I consider the latter to be borrowings from one or more linguistic varieties to which I give the collective name of Pyrenaic¹.

The observed differences between Pyrenaic and Proto-Basque are basically due to the conditioned aspiration of stops p, t, k into h, a sound change which I call Martinet's Law.  Therefore, I reconstruct a prehistoric stage of Basque before it which I name Early (Martinet's) Proto-Basque to differentiate it from Late (Mitxelena's) Proto-Basque. In this way, Pyrenaic and Early Proto-Basque look very similar.

Other isoglosses (mainly regarding to vocalism in the first syllable²) made me distinguish between two language groups, Cantabrian and Tyrrhenian. Some Tyrrhenian words seem to have reached Basque through Iberian, as part of its lexicon is from that origin. For example ui (B) 'pitch' < *uni, from an earlier **guni (PNC *kk’wVnV 'mastics, tar'), with assimilation to *buni and regular loss of the labial stop³. By contrast, Basque koipe, goipe 'oil' < *goni-pe (a compound from the same root) keeps the velar stop.

Very often, Cantabrian words have Tyrrhenian counterparts and viceversa. For example, Basque taket 'stake, wedge; dump, stupid' corresponds to Spanish zoquete 'block of wood; dumb, stupid'.
_____________

¹ This name was first suggested by my Italian colleague Marco Moretti, who leaded me in the earlier stages of my research. Pyrenaic is also the source of many substrate loanwords in Romance languages (specially Spanish) as well as part of the Iberian lexicon.

² In many words, Cantabrian has a/e while Tyrrhenian has o/u.
³ Although Bengtson identified this process, he wrongly attributed it to native Basque.

The magpie and the heron

There's a class of onomatopoeic roots *kVr- which designate several types of birds.

For example, Aragonese garza, Catalan garça, Italian gazza 'magpie' < *karkea is an Italoid loanword from PIE *ḱarhk-eh2- 'magpie' (Lithuanian šárka, Russian soróka)¹, akin to Turkic *KArga 'crow' and PNC *q’q’HVrVq’V 'a k. of bird (magpie; eagle-owl)'.

The homonymous Spanish garza 'heron'² is a loanword from Celtic *kor(x)sā 'heron, crane', probably through a Cantabrian intermediate *karsa. A similar root can be found in Latin ardea 'heron' < PIE *h1orhd-eh2- 'heron', akin to Turkic *Kordaj 'pelican; swan'.

Latin corvus 'raven' is a loanword from Afrasian *ɣurVb- 'crow, raven', and cornīx 'crow' is akin to Uralic *kOrnV 'raven'.
_____________
¹ Probably also Celtic *kerkā 'hen'.
² In Spanish, 'magpie' is urraca.

Spears and poles

Basque langa 'enclosure, rustical door; bar, catch (on a door or window)'¹ is related to Occitan/Catalan tanca id.², with a sound shift *t- > l- in Proto-Basque.

This is an Italoid substrate loanword whose etymology is IE *tengh-s- 'pole' > Latin tēmō 'steering-wheel; spear (of a cart)' and Old English þīsl 'wagon-pole, shaft'³. This IE root is a -n- infixed variant of *(s)teg- 'pole, post' (English stake), which in turn is related to PNC *dwɨq’(w)V: 'log, stump'.

This Vasco-Caucasian root is found as a Vasconic loanword in Spanish taco, Catalan tac 'block of wood, wedge; wooden stick', French taquet (diminutive form) 'block of wood, wedge', as well as in Basque haga 'pole, stake'⁴. The Tyrrhenian variant is found in Spanish tocón (augmentative form), Portuguese tôco 'stump' and also in Catalan soca, French souche 'stump', Old French choque 'log' and Spanish zoquete (diminutive form) 'block of wood; dumb, stupid'⁵, with assibilation of the initial dental.
___________
¹ Mitxelena derives it from Romance *planka 'plank, board'.
² Coromines proposes a derivation from a hypothetical Sorotaptic verb *tankō- 'to close'.

³ See Mallory & Adams (2006), p. 249.
⁴ There's also non-standard Basque tako 'circular piece of wood' and taket 'stake, wedge; dump, stupid'.

⁵ Coromines considered this word to be an arabism. What a zoquete!

Scabies and iron

English iron (and related forms in other Germanic languages) is a loanword from Celtic *īsarno- 'iron', whose origin is the IE adjective *H1ésH2r̥-no- 'bloody', derivated from *H1ésH2r̥- '(flowing) blood'.

Although nobody thought of it until now, *H1ésH2r̥-no- is also the origin of Spanish sarna 'scabies, mange', a pre-Romance word glossed by Latin authors. There're also Basque sarra 'rust, iron tartar'¹ and Spanish sarro 'tartar, plaque', with assimilation -rn- > -rr-.

The source of this and other loanwords is an IE substrate language called Italoid by the Indo-Europeanist Francisco Villar² and Sorotaptic by the Catalan linguist Joan Coromines³. According to them, Italoid shares isoglosses with Baltic and Italic in the IE dialectal cloud.

_____________

¹ Only found in the Biscayan dialect, which also has sarna.
² See Indoeuropeos y no indoeuropeos en la Hispania prerromana (2000).
³ Who has a quite different etymological proposal for this word.

The rock

Romance *rokka 'rock' (Italian rocca, French roche, Occitan ròca, Spanish roca, Portuguese rocha) is a Tyrrhenian loanword related to PNC *riqq’wA 'mountain, rock; cave'.

Basque arroka, harroka has a prothetic vowel also found in Gascon arròca¹. This is so because Basque doesn't allow rhotics at word-initial.
__________
¹ The Basque word is probably a borrowing from Gascon.